@earthling shouldn't we either use the "old way" or the "new way" depending if ipv6 is used or not ? Also could you please sign our ICLA and squash the commits ?
http://netty.io/s/icla

I'm not sure what you mean by 'old way' and 'new way'? It seemed to me like the same socket could be used for IPv4 and IPv6, so I just set both socket options.

I signed the ICLA and I'll try to squash the commits (I'm terrible with git). Should I open a new pull request for the squashed commit?

@earthling well I think we either need to use one or the other option set depending on ipv6 or ipv4. And we also need to fix this for the "get" operation as well.

@normanmaurer are you suggesting to check if the socket type is AF_INET or if IPV6_V6ONLY is true? What about the case when the socket type is AF_INET6 and IPV6_V6ONLY is false? would you not want to cover both ipv4 and ipv6 at the same time? Or are you suggesting we expose these separately in the Java API? Setting the IPv6 option on an IPv4 socket (or setting the IPv4 option on an IPv6 socket) seems harmless. I agree we need to fix the 'get' operation, I can fix that once we settle on an implementation for the 'set' operation.

@earthling yeah I would check the socket type. I think this should be good enough. Depending on the socket type we then use the correct opts for set and get.

Something like this then?

    if (socketType == AF_INET) {
        netty_unix_socket_setOption(env, fd, IPPROTO_IP, IP_TOS, &optval, sizeof(optval));
    }
    else {
        netty_unix_socket_setOption(env, fd, IPPROTO_IP, IP_TOS, &optval, sizeof(optval));
        netty_unix_socket_setOption(env, fd, IPPROTO_IPV6, IPV6_TCLASS, &optval, sizeof(optval));
    }

@earthling yeah but for the "else" you should only need:

        netty_unix_socket_setOption(env, fd, IPPROTO_IPV6, IPV6_TCLASS, &optval, sizeof(optval));

Yes, but then it wouldn't work on IPv4 packets sent via that socket.

A full accounting would look like:

    if (socketType == AF_INET) {
        netty_unix_socket_setOption(env, fd, IPPROTO_IP, IP_TOS, &optval, sizeof(optval));
    }
    else if (IPV6_V6ONLY) {
        netty_unix_socket_setOption(env, fd, IPPROTO_IPV6, IPV6_TCLASS, &optval, sizeof(optval));
    }
    else {
        netty_unix_socket_setOption(env, fd, IPPROTO_IP, IP_TOS, &optval, sizeof(optval));
        netty_unix_socket_setOption(env, fd, IPPROTO_IPV6, IPV6_TCLASS, &optval, sizeof(optval));
    }

@earthling makes sense... But what we do with the get ? I guess just "query" on in the else should be good enough. Agree ?

Also please use the coding style we use in our .c files and update the commit message to match our template:

http://netty.io/wiki/writing-a-commit-message.html

Okay, will do. Should I make a new pull request with all the changes?

@earthling you can also just update this one...

@normanmaurer Thinking about the 'get' operation... I think it's going to be okay to just use the IP_TOS value. getsockopt returns this value correctly for AF_INET6 sockets (even ones when IPV6_V6ONLY is true). I did change the 'set' operation to not raise an exception if we try to set IPV6_TCLASS on an AF_INET only socket.

@earthling sounds good. Let me know once you are done.

@normanmaurer I think I'm done. Thanks for working this out with me.

@earthling so the get operation not need any changes ?

I don't believe so, no. For all socket types, the value set for IP_TOS is returned with getsockopt, and we set IP_TOS for all socket types.

@earthling thanks!

Squashed your commits, adjusted the commit message and cherry-picked into 4.0 (ae470e3) and 4.1(c205e2b)

@normanmaurer Thank you!